The statistical outcomes for this experiment are described before the

outcomes in Experiment 3 are discussed. In Experiment 4, repeated hits are targets, and studied words are non-targets; the reverse from the target/non-target designation in previous experiments, (as well as subsequent) in this thesis.

Hits, correct rejections and repeated hits

Figure 6.6 is relevant to the following statistical analysis and shows the ERPs elicited by hits, correct rejections and repeated hits at electrode sites F3/F4, C3/C4, P3/P4, Fz, Cz and Pz. The figure shows that from approximately 300 ms onwards the ERPs

elicited by targets are more positive-going than those elicited by non-targets and new items. This relatively greater positivity peaks at 600 ms and by the end of the

recording epoch has attenuated with non-targets becoming marginally more positive- going, particularly at central and posterior midline sites. The old/new effect for non­ targets is relatively small in magnitude from 300 ms until 600 ms where non-targets become more positive-going, particularly at central and right frontal sites.

300-500 ms: Midline analyses:

The initial midline analysis revealed a main effect of category (F(1.4, 24.5) = 10.76, p<.001). The follow-up analyses revealed that the ERPs evoked by hits and repeated hits were reliably more positive-going than those elicited by correct rejections (F(l,17) = 7.15, p<.05 and F(1,17) = 15.16, p<.001, respectively). ERPs evoked by repeated hits were also more positive-going than those elicited by hits (F(l ,17) = 6.95, p<.05).

Global analyses:

The second set of analyses for this time window including the data from lateral scalp locations revealed a main effect of category (F(1.6, 27.2) = 14.28, p<.001), as well as an interaction between category and site (F(1.5, 26.1) = 6.66, p<0.01). The follow-up analyses comprised all possible paired contrasts of the three categories and the results are shown in Table 6.3. As for previous comparable tables, only those terms for which outcomes were reliable in at least one paired contrast are shown.

Table 6.3. The outcomes of the paired contrasts between the mean amplitudes

associated with hits, correct rejections and repeated hits for the global analyses over the 300-500, 500-800 and 800-1100 ms time windows. Nomenclature as for Table 5.2, page 137.

300-500 ms

df RHit vs CR RHit vs Hit CR vs Hit

CC 1,17 24.73*** 6.92* 10.15** CCx SI 2,34 9.73(53)*** ns 7.18(67)* CC x AP 2,34 6.52(.79)** ns ns 500-800 ms CC 1,17 39.01*** 32.26*** ns CCx SI 2,34 39.83(54)*** 31.16(.56)*** 5.67(.65)* 800-1100 ms CC 1,17 10.10** ns 11.53**

The contrast between repeated hits and correct rejections revealed an interaction between category and site, which was due to the relatively greater positivity for repeated hits being largest at sites closest to the midline. There was also an interaction between category and AP, which reflected the fact that the old/new effects were largest at central sites. Repeated hits were also more positive-going than hits in this time window. For the contrast between hits and correct rejections there was a significant interaction between category and site which reflected the fact that the relatively greater positivity for hits was largest at midline sites.

500-800 ms: Midline analyses:

The initial midline analysis revealed a main effect of category (F(1.3, 22.5) = 35.78, p<.001). The follow-up analyses revealed ERP old/new effects for hits and repeated hits (F(l,17) = 8.42, p<.01 and F(l,17) = 43.58, p<.001, respectively). Repeated hits were also reliably more positive-going than hits (F( 1,17) = 33.18, p<.001).

Global analyses:

The second set of analyses for this time window revealed a main effect of category (F(1.5, 25.0) = 31.19, p<.001), as well as an interaction between category and site (F(1.6, 27.0) = 31.65, p<.001). The follow up paired contrast results are shown in Table 6.3.

The contrast between repeated hits and correct rejections revealed an interaction between category and site which was due to the relative greater positivity for

repeated hits being largest at midline sites. For the contrast between correct

rejections and hits there was an interaction between category and site because of the relative greater positivity for hits at midline sites. The final contrast between hits and repeated hits revealed an interaction between category and site that came about because the relatively greater positivity for repeated hits was maximal at midline sites.

800-1100 ms: Midline analyses:

The initial analysis revealed no significant effects; no follow-up analyses were conducted.

Global analyses:

The second set of analyses for this time window revealed a main effect of category (F(1.4, 23.4) = 7.34, p<.01). The follow-up contrast results are shown in Table 6.3. These revealed reliable positive-going old/new effects for hits and repeated hits. There were no significant effects in the contrast between hits and repeated hits.

Mid-frontal ERP old/new effects:

The directed analysis of the mid-frontal ERP old/new effects involved the mean amplitudes for hits, repeated hits and correct rejections at F3, Fz and F4 in the 300- 500 ms time window. Figure 6.7 is relevant to the following statistical analyses and shows the ERPs elicited by the three classes of items at these electrode sites. The figure shows that from approximately 300 ms repeated hits are more positive-going than hits and correct rejections, with minimal differences between hits and correct

rejections. This differentiation at frontal sites continues throughout the recording epoch.

The analyses revealed a main effect of category only (F(1.7, 28.3) = 14.25, p<.001). Paired contrasts within the experiment demonstrated that hits (F(l,17) = 5.80, p<.05) as well as repeated hits (F(l,17) = 20.03, p<.001) were both more positive-going than correct rejections at these mid-frontal electrode sites. Repeated hits were also more positive-going than hits (F(l,17) = 11.42, p<.005) with this relative greater positivity maximal at Fz (F(2.0, 33.4) = 3.70, p<.05). This can clearly be seen in Figure 6.3, which shows a topographic map illustrating the scalp distribution of the differences between activity evoked by repeated hits and that evoked by hits over the 300-500 ms time window.

Left-parietal ERP old/new effects:

The parietal old/new effects directed analysis involved the mean amplitudes for hits, repeated hits and correct rejections at P3, P5, P4 and P6 in the 500-800 ms time window. Figure 6.8 is relevant to the following statistical analyses and shows the ERPs elicited by the three classes of items at electrode sites P3/4 and P5/6. The figure shows that in the critical 500-800 ms time window repeated hits are more positive-going than hits and correct rejections, with minimal differences between hits and correct rejections. This greater relative positivity for repeated hits is maximal at P3/4.

The further analysis revealed a main effect of category (F(1.4, 24.0) = 31.69, p <.001). This was followed up by all possible paired contrasts, and the outcomes of

these showed that ERPs evoked by repeated hits were reliably more positive-going than those evoked by correct rejections (F(l,17) = 39.76, p<.001) and this relative greater positivity was maximal at midline sites (F(2.1, 35.7) = 5.04, p<.05). The greater positivity for repeated hits relative to correct rejections is illustrated in Figure 6.3 (right-hand lower panel) which shows a topographic map of the differences in Scalp activity between repeated hits and correct rejections, and demonstrates clearly that the difference between the two conditions is maximal over central scalp regions. Repeated hits were also reliably more positive-going than hits (F(l,17) = 31.38, p<.001). The contrast between the ERPs evoked by hits and by correct rejections revealed no reliable effects over this time window.

Discussion

For experiments 1 and 2 parietally distributed ERP old/new effects were observed for correct responses to both studied and repeated test words. This is in stark contrast to the findings of Dywan et al. (1998, 2001, 2002) who utilised the same paradigm. Within the three publications of Dywan and colleagues there is little evidence for a relatively greater positivity for correct responses to repeated test words in

comparison to correct rejections at parietal locations in the 500-800 ms time window.

One account for the disparities across studies arises from the fact that, in experiments 1 and 2, the ratio of studied words to repeated test words was 2:1, in comparison to the 1:1 ratio in the studies of Dywan et al. This item imbalance is relevant because of the P300 ERP component. The P300 (P3b) potential is particularly sensitive to the relative probabilities of classes of stimuli, as well as their task-relevance (Donchin & Coles, 1988; Polich, 2007). By this account, the greater relative positivity for

repeated test words relative to new words in experiments 1 and 2 but not in the work of Dywan et al. is simply a consequence of a larger P300 being elicited by the repeated test words in experiments 1 and 2 (see data for Pz shown in Figures 5.1 and 5.7 in Chapter 5).

The motivation for Experiment 3 was to assess the impact on the left-parietal ERP old/new effects for hits and repeated hits of changing the probabilities of occurrence of these items to match the 1:1 ratio in the Dywan et al. studies. In the following sections there will be a discussion of the ERP findings of differences according to condition in the 300-500 ms time window and how these are relevant to the functional interpretation of the mid-frontal ERP old/new effect. This will be

preceded by a discussion of the critical behavioural and ERP experimental findings relevant to the principal focus in these two experiments: the left-parietal ERP old/new effects that were obtained

Left-parietal ERP old/new effects

The 500-800 ms epoch is one in which left-parietal ERP old/new effects are typically assessed (Friedman & Johnson, 2000; Mecklinger, 2000; Wilding & Sharpe, 2003), and in this time wndow the directed analyses showed reliable ERP old/new effects for targets in both experiments, and for non-targets in Experiment 3 only (cf.

Bridson, Fraser, Herron, & Wilding, 2006). Targets were denoted as studied words in Experiment 3, and as repeated test words in Experiment 4. There was a trend for the left-parietal ERP old/new effect for targets in Experiment 3 to be larger than the effect for non-targets.

The pattern of data obtained in Experiment 3 is the same as that obtained in

experiments 1 and 2, and if the magnitude of left-parietal ERP old/new effects is an index of the extent to which recollection was engaged, then the findings are

consistent with the view that participants did not prioritise recollection of studied words over repeated test words any more so in Experiment 3 than they did in experiments 1 and 2.

The change in the probability of occurrence of the repeated test words relative to studied words to match that used in the Dywan et al. studies did not, therefore, attenuate the parietal ERP old/new effect for the repeated test items. These findings lead to the conclusion that the disparity in findings across studies is most likely not due to a probability imbalance, which might have led to a relatively greater

contribution of the P300 to parietal old/new effects for repeated test items in

experiments 1 and 2 than in the studies reported by Dywan et al. (1998, 2001, 2002).

This conclusion draws added support from the fact that, at posterior scalp locations in Experiment 3 there was some degree of left hemisphere lateralisation for both old/new effects, (illustrated in Figure 6.3) although this was statistically reliable. In so far as the left-parietal ERP old/new effect has a left hemisphere bias that is not characteristic of the P300 (Azimian-Faridani & Wilding, 2006: Rugg & Allan, 2000), then the data are inconsistent with a P300 probability imbalance account.

There are two additional and compelling reasons, moreover, for rejecting the response probability account of the data. First, ERP old/new effects for words

tasks where the words encoded under these conditions were designated as targets, but the proportions of targets, non-targets and new test words were equal (Herron & Rugg, 2003b).

Second, Herron et al. (2003) conducted an ERP study in which old/new recognition memory judgments were required, and where, across blocks, the ratios of new to old test items were 3:1,1:1, and 1:3. Over the 500-800 ms period at left-parietal scalp sites, the positive-going ERP old/new effects that were obtained were of equal magnitude across these different ratios of old and new words. The 1:3 old/new ratio is directly comparable to the ratio of targets to the other (combined) classes of test stimuli in experiments 1 and 2 (for prior work relevant to the invariance of parietal old/new effects across other rations of old and new test items, see Friedman, 1990; Rugg & Nagy, 1989; Smith & Guster, 1993). These observations argue against a response probability account of the data in these experiments, and further support for this view stems from the fact that qualitatively similar findings to those in

Experiment 3 were found in experiments 1 and 2 where studied words designated as targets comprised 40% of the test stimuli (see Chapter 5). The remaining possible reasons behind the disparities between the findings in experiments 1, 2 and 3 and those of Dywan et al. (1998, 2001, 2002) will be returned to later.

Target accuracy

By virtue of having the repeated test items as targets the mean level of target accuracy for Experiment 4 (0.82) was higher than in any case when studied words were targets (see the previous three experiments). This is most likely due to the shorter lag between first and second presentations of the repeated test items

compared to the studied items. This experiment is also unlike the previous three in that, at parietal locations in the 500-800 ms time window, there was a complete absence of a left-parietal ERP old/new effect for the non-targets (studied items). When taken together with the previously presented experiments, the differential attenuation of non-target parietal old/new effects replicates findings in four other cases (Dzulkifli et al., 2006; Dzulkifli & Wilding, 2005; Herron & Rugg, 2003a; Herron & Rugg, 2003b; Wilding et al., 2005). In keeping with the findings in each of these pairs of experiments, the attenuation of the ERP old/new effects for non-targets that is reported in Experiment 4 occurred in an experiment in which the likelihood of a correct target judgment was much higher than in previous studies where there had been no observed attenuation (see experiments 1, 2 and 3). The findings of

Experiment 4 therefore support the view that participants adopt strategies in

exclusion tasks, whereby the extent to which recollection of information of targets is likely determines the extent to which recollection of information about non-targets will be prioritised (Herron & Rugg, 2003b).

The levels of target accuracy across experiments 1, 2 and 3 are comparable to those reported in previous exclusion studies (cf. Dzulkifli et al., 2006; Dzulkifli &

Wilding, 2005; Herron & Rugg, 2003a, 2003b; Wilding et al., 2005) where some degree of attenuation of non-target old/new effects has been reported. Why, therefore, is there no comparable degree of attenuation in experiments 1-3? As described previously, it may be the case that, in comparison to the tasks demands in standard study-test exclusion tasks, and the version of the paradigm used in

Experiment 4, in experiments 1-3 participants experience a greater degree of difficulty in controlling recollection of the repeated test items because of the short

lag between their first and second presentation. Therefore, the previous findings may be due to relatively lower degrees of difficulty in controlling the recollection of the non-target categories employed in those experiments. In order to assess this, one way is to provide a greater incentive to process selectively information about targets, by increasing the likelihood of recollecting information about targets. The issue of target/non-target accuracy levels in this paradigm will be addressed in Experiment 5.

Mid-frontal ERP old/new effect

Focusing now on the data acquired from 300-500 ms post-stimulus at anterior scalp locations; does the data contribute to one interpretation offered for findings in previous research (Curran, 2000), which is that frontally distributed ERP old/new differences in this time window are related to familiarity?

The directed analysis of the mid-frontal ERP old/new effect confirmed that there were reliable ERP old/new effects for both studied and repeated test items in both experiments. Whilst there were no differences between the two classes of old item and misses in Experiment 3, misses were more positive-going than correct rejections, as indicated by an interaction between category and site, which reflected a greater relative positivity for misses at Fz only. This pattern of differences between

conditions is similar to that in experiments 1 and 2 (see Figure 6.5 for a bar graph of the mean amplitudes of the old/new effects for hits, misses and repeated hits across frontal sites in the 300-500 ms time window, which is directly comparable to the data shown in Figure 5.8 in Chapter 5). The data is in line with the proposal that an

anteriorly distributed modulation occurring in this time window is an index of familiarity (for a comprehensive review, see Curran et al., 2006), although, because

of the absence of reliable differences between the ERPs elicited by misses and by the two classes of hit, it is not as statistically compelling as in the earlier experiments.

Furthermore, the distributions of the old/new effects for repeated hits and misses within this time window (see Figure 6.3) show maxima with a somewhat more posterior loci than would be anticipated if they were solely reflecting mid-frontal ERP old/new effects. This may be due to the coexistence of two distinct processes in this time period. It has been proposed that two functionally distinct ERP old/new effects, with fronto-central and parietal scalp distributions respectively, are evident in the 300-500 ms post-stimulus time-window (Azimian-Faridani & Wilding, 2006; Rugg et al., 1998). Rugg et al. (1998) linked the posterior effect with implicit memory, since in their study the effect differentiated only the old/new status of test items and not the accuracy of memory judgments. The data shown in Figure 4 are consistent with this account, as at posterior sites correct responses to studied words and repeated test words, as well as incorrect responses to studied words, are more positive-going than correct rejections. The ERPs at the posterior locations thus index the old/new status of the test items but do not predict the accuracy of memory judgments. This pattern of data fulfils one criterion for being an index of implicit memory, but in this study, as in all prior studies to date where this has been

demonstrated, the effect has not been accompanied by a behavioural index of implicit memory such as priming (Azimian-Faridani & Wilding, 2006).

Concluding remarks

The directed analysis of the mid-frontal ERP old/new effects confirmed that within both experiments there were reliable old/new effects for both classes of old items and

that misses were also more positive-going than correct rejections at midline anterior sites only for Experiment 3. Furthermore, in Experiment 4, repeated hits were reliably more positive-going than hits. The pattern of data obtained here is in line with the proposal that an anteriorly distributed modulation occurring in the 300-500 ms time window indexes familiarity (Rugg et al, 1998), because at anterior sites only, the ERPs elicited by correct responses to targets and non-targets were more positive-going than those elicited by correct rejections. This pattern of data is consistent with the view that frontally distributed ERP old/new differences (appearing between 300 and 500 ms) are related to familiarity (Curran, 2000).

For Experiment 3 there were reliable parietal ERP old/new effects for both studied and repeated test items, with the same pattern of differences as for experiments 1 and 2. For Experiment 4, where the target/non-target designation was reversed, there were reliable parietal ERP old/new effects for the repeated test items only.

Experiment 4 was primarily designed in order to formally assess the memorability of